Albert-Jan N. Yzelman

What's new

• With an extension to the red-blue pebbling game we combine vertical I/O with horizontal data movement models commonly applied in parallel computing. This culminates in a model that captures trade-offs between work, communication, and memory, thus combining classic work in theoretical computer science with the state-of-the-practice in high-performance computing. It attempts to further advance our past investigations in arbitrary DAG scheduling over increasingly realistic models (see here from a theory perspective and here from a practical perspective) by accounting for limited memory in arbitrary system hierarchies (e.g., caches). One may also view this as extending our previous contributions from BSP-type models to that of the more recently proposed Multi-BSP. For more details, the pre-print is available and has just been accepted for presentation at SIROCCO 2025, Delphi, Greece.
• Our work on a fast distributed-memory (dense) tensor–vector multiplication (TVM) that remains oblivious to contraction mode, splitting dimension, and order of the tensor has been accepted with Future Generation Computer Systems. The work includes performance and scalability studies over multiple architectures, application of the method to the higher-order power method, and mixed-precision TVMs. Update: the paper is published and a free temporary publisher link is available.
• My text on Humble Heroes (citation) has left press. Humble here refers to the notion of the humble programmer, a concept first coined by Edsger Dijkstra, who urged us to approach programming while respecting the intrinsic limitations of the human mind. Today, few programming models are humble. Even fewer are both humble and scalable. Almost none are humble, scalable, and high performance.
  The ever-increasing scale and complexity of computing systems – inspired by ever-increasing scales of problems and data – however, requires ever-more programmers, each with ever-deeper expertise, to write programs that require ever-higher scalability and performance.
  In a bid to resolve this juxtaposition, Humble Heroes argues many humble programming models can be built on top of a few foundational ones, while retaining high performance and optimal scalability– and demonstrates this is feasible by introducing and evaluating ALP/Pregel, freely available on GitHub.

Contact info

Overview

• Publications
• Presentations
• ALP/Pregel & ALP/GraphBLAS (gitee), MulticoreBSP, and other software
• A short biography, and an even shorter one